Trajectory tracking control method and system for high-load omnidirectional AGV

Abstract

The invention provides a trajectory tracking control method and system for a high-load omni-directional AGV, the omni-directional AGV is driven by a plurality of steering wheel mechanisms which are independently steered and driven, and the trajectory tracking control method is characterized by comprising the following steps: acquiring vehicle state information and expected trajectory information of the omni-directional AGV in real time; on the basis of the vehicle state information and the expected track information, core parameters of a model prediction controller are optimized online through a chaos simulated annealing algorithm, and the core parameters comprise a state weight matrix, a control increment weight matrix, a prediction time domain and a control time domain; wherein the model prediction controller performs state prediction based on a longitudinal-transverse decoupling dynamic model adaptive to a high-load working condition; performing rolling horizon optimization through the model prediction controller by utilizing the optimized core parameters, and solving to obtain an optimal control increment; and converting the optimal control increment into a driving instruction for a multi-steering-wheel mechanism, and controlling the omnidirectional AGV to track an expected trajectory.

Description

Technical Field

[0001] This invention belongs to the technical fields of industrial automation logistics and power construction equipment, specifically relating to a trajectory tracking control method and system for high-load omnidirectional AGVs. It is mainly applied to high-voltage cable handling scenarios, and is particularly suitable for high-load, high-mobility handling equipment in narrow and confined spaces such as substations, underground utility tunnels, and cable trenches. Background Technology

[0002] High-voltage cable handling operations often involve narrow and confined spaces such as substations, underground utility tunnels, and cable trenches. Each cable can weigh several tons, placing stringent demands on the load-bearing capacity, mobility, trajectory accuracy, and operational safety of the handling equipment. Automated Guided Vehicles (AGVs), as core equipment in the intersection of industrial automation logistics and power construction equipment, have become the preferred solution in this scenario due to their autonomous movement and precise positioning capabilities. The drive structure design and intelligent control strategy of their chassis system directly determine operational efficiency, safety, and adaptability.

[0003] Existing high-voltage cable handling AGV chassis generally adopt a multi-steering wheel drive structure. Among them, the four-steering wheel symmetrical layout has become the mainstream design for operation in confined spaces because it can achieve omnidirectional movements such as translation, rotation, and path following. Some solutions simplify the four-steering wheel model to a two-steering wheel model based on the Ackermann steering principle to reduce control complexity, thereby improving control feasibility by simplifying the dynamic relationship. However, this simplification often involves a trade-off between control accuracy and adaptability to complex movements. In terms of control strategies, basic path following mostly relies on PID control algorithms. This algorithm has a simple structure and low engineering implementation cost. It adjusts the steering wheel deflection angle and speed parameters by detecting trajectory deviation and rate of change to achieve basic path following function. However, in the complex dynamic environment of high-voltage cable handling, its anti-interference ability is limited, and it is difficult to cope with the effects of sudden disturbances in confined spaces, load fluctuations caused by high loads, and uneven road surfaces. In advanced control schemes, Model Predictive Control (MPC) is widely used in AGV trajectory tracking optimization because it can predict the system state in the future time domain and effectively adapt to the requirements of multi-constraint systems. However, traditional MPC schemes do not introduce a global optimization mechanism, and the setting of core parameters such as objective function weights and prediction time domains largely depends on engineering experience, lacking adaptive adjustment capabilities and easily getting trapped in local optima. Furthermore, they suffer from insufficient stability in complex trajectory tracking scenarios such as sine waves. Regarding fault-tolerant control, existing technologies mostly address hardware failures such as steering wheel failure by establishing differential steering models. However, they lack effective adaptation and compensation mechanisms for dynamic disturbances such as changes in road slope, fluctuations in air resistance, and slight shifts in load distribution. The robustness of the system under complex operating conditions needs further improvement.

[0004] Several related technical solutions have been disclosed. For example, some patents propose using a four-steering wheel mechanism combined with a PID controller to decompose the steering wheel motion through an instantaneous rotation center to reduce sliding friction and improve the stability of basic path tracking. Another solution establishes a kinematic and dynamic model based on Ackermann steering theory and performs linearization processing, designs an MPC objective function and constrains velocity and angular velocity to achieve trajectory tracking control. Yet another technology simplifies the four-steering wheel model to a two-steering wheel model, and uses a differential drive controller to adjust the wheel speed difference to address the fault-tolerant control requirements in steering wheel failure scenarios. A review article on the application of MPC in autonomous ground vehicles also points out that although MPC technology can better meet the predictive decision-making needs of AGVs, it still faces common problems in practical engineering applications, such as high computational complexity, communication delays in multi-AGV collaborative operations, and the contradiction between low-cost hardware and real-time control requirements. These problems are particularly prominent in high-load AGV scenarios due to the more complex dynamic models.

[0005] Despite some progress in AGV chassis design and control, several shortcomings remain for the special working conditions of high-voltage cable handling. Regarding control accuracy and environmental adaptability, PID control exhibits significant trajectory tracking deviations in confined spaces and under high load conditions, while traditional MPC, lacking global optimization capabilities, suffers from poor stability in complex path tracking, failing to meet the high-precision trajectory following requirements of high-voltage cable handling. The conflict between computational efficiency and real-time performance is particularly pronounced; the complexity of the dynamic model increases significantly under high load scenarios, leading to slower convergence speeds in traditional MPC's rolling optimization, delayed control command response, and excessively long obstacle avoidance time, making it impossible to promptly avoid sudden obstacles in confined spaces. Furthermore, the centroid shift problem caused by high loads has not been adequately addressed, and most dynamic models have not been fully optimized. The decoupling optimization of longitudinal and lateral directions leads to excessive lateral acceleration during turning, which can easily cause rollover or trajectory deviation risks, making it difficult to ensure the stable handling of high-voltage cables weighing several tons. The fault tolerance and optimization mechanisms are not perfect. The key parameters of MPC rely on experience to set and lack adaptive adjustment capabilities. Moreover, the fault tolerance control is mostly limited to the scenario of steering wheel failure, and the coverage of various dynamic disturbances is insufficient, resulting in limited system robustness under complex working conditions. In addition, the existing solutions often have problems such as large path deviation and long recovery time during obstacle avoidance. In narrow spaces, they are prone to collisions with pipe walls and surrounding equipment, making it difficult to meet the stringent safety handling requirements of high-value goods such as high-voltage cables. Summary of the Invention

[0006] To address the shortcomings and deficiencies of existing technologies, this invention provides a trajectory tracking control method, system, and device for high-load omnidirectional AGVs. The core of this solution lies in first acquiring the AGV's vehicle status and desired trajectory information in real time; then, based on a longitudinal-lateral decoupled dynamic model specifically adapted for high-load conditions, using a chaotic simulated annealing algorithm, online optimization is performed on the model predictor's core parameters, including the state weight matrix, control increment weight matrix, prediction time domain, and control time domain; subsequently, rolling time domain optimization is performed using the optimized parameters to solve for the optimal control increment; finally, this increment is converted into drive commands for the multi-steering wheel mechanism to achieve precise trajectory tracking.

[0007] The innovative design points of this invention are mainly reflected in three aspects. First, at the model level, a composite dynamic model including lateral attitude, longitudinal resistance, and vibration disturbance compensation is constructed. This model not only decouples longitudinal and lateral dynamics but also compensates for road disturbance forces in real time through virtual suspension equivalent parameters, significantly improving the model prediction accuracy under high loads and complex road conditions. Second, at the control optimization level, a chaotic simulated annealing global optimization algorithm is innovatively combined with model predictive control. Through steps such as chaotic sequence initialization and Metropolis criterion for accepting new solutions, the key parameters of the MPC controller are dynamically optimized online, effectively avoiding the problems of fixed parameters and easy getting trapped in local optima in traditional methods, and enhancing the system's adaptability and tracking stability under different working conditions. Third, at the system robustness and coordination level, this method integrates an encoder feedback-based adaptive compensation mechanism for steering wheel wear and a multi-parameter monitoring fault tolerance strategy, and supports collaborative operation and role switching of multiple AGVs through wireless communication. Meanwhile, the supporting system adopts a hardware architecture that includes dual-redundant sensing and control, hierarchical power supply and thermal management, which together ensure the safe, efficient and high-precision reliable operation of the high-load AGV in narrow and complex industrial scenarios from both hardware and software perspectives.

[0008] The present invention specifically employs the following technical means:

[0009] A trajectory tracking control method for a high-load omnidirectional AGV, wherein the omnidirectional AGV is driven by multiple independently steering and driving steering wheel mechanisms, comprising:

[0010] Real-time acquisition of vehicle status information and desired trajectory information of the omnidirectional AGV;

[0011] Based on the vehicle state information and the desired trajectory information, the core parameters of the model predictive controller are optimized online using the chaotic simulated annealing algorithm. The core parameters include the state weight matrix, the control increment weight matrix, the prediction time domain, and the control time domain. The model predictive controller performs state prediction based on a longitudinal-lateral decoupled dynamic model adapted to high-load conditions.

[0012] Using the optimized core parameters, the model predicts the controller to perform rolling time-domain optimization, and the optimal control increment is obtained by solving the problem.

[0013] The optimal control increment is converted into a drive command for the multi-steering wheel mechanism, controlling the omnidirectional AGV to track the desired trajectory.

[0014] Furthermore, the longitudinal-lateral decoupled dynamic model includes a lateral two-degree-of-freedom model for describing changes in vehicle attitude, a longitudinal resistance model for calculating load resistance under high load conditions, and a vibration disturbance compensation sub-model. The longitudinal resistance model covers acceleration resistance, air resistance, slope resistance, and rolling resistance. The vibration disturbance compensation sub-model calculates road disturbance force based on the equivalent stiffness and equivalent damping of the virtual suspension, combined with real-time acquired vertical displacement and vertical velocity, and outputs it as a compensation term to correct the model.

[0015] Furthermore, the optimization process of the chaotic simulated annealing algorithm includes: generating a chaotic sequence using a Logistic mapping and mapping it to a feasible solution for the core parameters of the model predictive controller; calculating the initial temperature based on the extremum of the objective function of the initial solution and the preset initial acceptance probability; determining whether a new solution is acceptable using the Metropolis criterion; decreasing the temperature with the number of iterations according to a preset decay coefficient; and outputting the optimal core parameters when the number of iterations reaches a preset iteration threshold or the rate of change of the objective function meets the convergence condition. The prediction time domain and control time domain of the model predictive controller are dynamically configured according to the operating conditions, and its objective function includes a state tracking error term, a control increment constraint term, and a fault-tolerant relaxation term.

[0016] Furthermore, the process of converting the optimal control increment into drive commands includes: based on the Ackermann steering principle, calculating the instantaneous center of velocity through virtual dual steering wheel modeling, deriving the target deflection angle and drive speed of each steering wheel, ensuring the motion consistency of multiple sets of symmetrically arranged steering wheels, and avoiding sliding friction.

[0017] Furthermore, it also includes adaptive compensation and fault tolerance steps for steering wheel wear: a wear assessment model is established based on the cumulative value of the angle deviation fed back by the encoder and the running time. When the wear level exceeds or reaches the preset wear level threshold, the state weight matrix of the model predictive controller is adjusted and the steering wheel deflection angle command is corrected. The steering wheel motor current, power supply voltage and equipment temperature are monitored in real time. When the monitored data exceeds or reaches the preset monitoring threshold, a graded early warning mechanism is triggered. When steering wheel failure or voltage abnormality is detected, the fault tolerance mode is activated and the control command of the healthy steering wheel is reallocated.

[0018] Furthermore, the control module of the omnidirectional AGV supports a multi-AGV collaborative interface, which realizes the position synchronization, load distribution and path coordination of multiple AGVs through wireless communication protocols, and supports the switching of master and slave AGV roles. The master AGV is responsible for path planning, and the slave AGV receives synchronization control commands.

[0019] And, a trajectory tracking control system for a high-load omnidirectional AGV, the omnidirectional AGV being driven by multiple independently steering and driving steering wheel mechanisms, comprising:

[0020] The perception module is used to collect the vehicle status information and desired trajectory information of the omnidirectional AGV in real time.

[0021] The control module is configured to optimize the core parameters of the model predictive controller online using a chaotic simulated annealing algorithm based on the vehicle state information and the desired trajectory information. The core parameters include a state weight matrix, a control increment weight matrix, a prediction time domain, and a control time domain. The model predictive controller performs state prediction based on a longitudinal-lateral decoupled dynamic model adapted to high-load conditions, and then performs rolling time domain optimization using the optimized core parameters to solve for the optimal control increment.

[0022] The drive module is used to receive the optimal control increment output by the control module, convert it into drive instructions and execute them to control the omnidirectional AGV to track the desired trajectory.

[0023] The power supply module is used to provide stable power to the sensing module, control module and drive module;

[0024] The fault diagnosis module is used to monitor the operating status of each module in real time.

[0025] Furthermore, the control module executes the trajectory tracking control method as described in claim 1 to achieve a closed-loop control process encompassing vehicle state information acquisition, model predictive controller parameter optimization, optimal control increment solution, and drive command generation.

[0026] Furthermore, the perception module collects vehicle status information through lidar, vision sensors, and inertial measurement units. The collected data is preprocessed by hardware-level low-pass filtering and Kalman filtering before being output. The perception module and control module adopt a dual redundancy design.

[0027] Furthermore, the power supply module adopts a multi-voltage level hierarchical power supply architecture, with dual-line redundancy configured for key modules. During braking, electrical energy is recovered through an energy recovery unit. The power supply module also integrates a thermal management unit to maintain stable operating temperature.

[0028] And a computer device including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the method described above.

[0029] A non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method described above.

[0030] Compared with the prior art, the present invention and its preferred embodiments have at least the following beneficial effects:

[0031] This invention effectively addresses the shortcomings of traditional control algorithms, such as being prone to getting trapped in local optima and relying on empirical parameter settings, through deep integration and optimization of Chaotic Simulated Annealing (CSA) and Model Predictive Control (MPC). It significantly improves control robustness and trajectory tracking accuracy under complex working conditions, enabling the system to better adapt to the complex paths and dynamic environments required for high-voltage cable handling. Addressing the core requirements of high-load scenarios, a customized longitudinal-lateral decoupled dynamic model combined with a vibration compensation mechanism accurately captures the impact of centroid shift and road surface disturbances on the motion state, greatly enhancing operational stability under high loads and effectively suppressing the risks of rollover and trajectory deviation. The four-steering wheel cooperative control strategy based on the Ackerman steering principle ensures the consistency of multi-steering wheel motion, reduces sliding friction, balances equipment maneuverability and operational reliability, and is more suitable for operations in narrow and confined spaces.

[0032] Meanwhile, the introduction of a full-link fault tolerance mechanism and adaptive compensation logic for steering wheel wear broadens the fault tolerance coverage, enhances the system's continuous operation capability under complex interference and long-term operation, and ensures the safety and continuity of high-value cargo handling. The distributed sensing, centralized control architecture, and hierarchical power supply and energy recovery design adopted by the hardware system not only ensure the high efficiency of data transmission and command execution, but also optimize energy utilization efficiency, extend the driving range, and reduce long-term operating costs. Overall, this invention achieves deep synergy between hardware support, algorithm decision-making, and model prediction, significantly improving control accuracy, real-time response, operational stability, robustness, and engineering practicality, and can fully meet the stringent requirements of high-voltage cable handling in scenarios such as substations and underground utility tunnels. Attached Figure Description

[0033] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments:

[0034] Figure 1 This is a hardware and algorithm collaborative framework diagram of an embodiment of the present invention. Detailed Implementation

[0035] In the following, specific embodiments of this application will be described in detail with reference to the accompanying drawings. Based on these detailed descriptions, those skilled in the art will be able to clearly understand and implement this application. Without departing from the principles of this application, features from various embodiments can be combined to obtain new implementations, or certain features from some embodiments can be substituted to obtain other preferred implementations.

[0036] To address the shortcomings and deficiencies of existing technologies, this invention aims to specifically resolve key technical challenges faced by high-voltage cable handling AGV chassis in confined spaces and under 5-ton heavy load conditions. Specifically, this includes: improving trajectory tracking accuracy for complex paths such as sine waves, controlling the average trajectory deviation to within 0.05m to meet precise handling requirements; optimizing the convergence speed of the MPC algorithm to ensure obstacle avoidance response time ≤1s, meeting the requirement for rapid response to sudden obstacles; and controlling the maximum lateral acceleration during turning to 1.2m / s² through dynamic decoupling and centroid offset compensation. 2 Within this range, the risk of instability under high load is suppressed; a global optimization algorithm is introduced to achieve adaptive adjustment of MPC parameters, improving the system's adaptability to dynamic disturbances such as road slope and air resistance; at the same time, the obstacle avoidance path offset is reduced to ≤0.5m and the path recovery time is shortened to ≤1s, ensuring the safety and continuity of the handling process and avoiding collisions with the surrounding environment.

[0037] (I) Overall Design

[0038] like Figure 1 As shown, the systematic design scheme provided by this invention is supported by hardware support, algorithm decision-making, data closed loop, and energy guarantee. It deeply integrates with the full-process collaborative logic of perception-computation-execution-feedback-optimization, realizing efficient linkage and functional complementarity between modules. It mainly includes:

[0039] 1. Perception Module: Includes 1 LiDAR, 2 vision sensors, 1 IMU (Inertial Measurement Unit), and 1 data fusion board, used to collect raw data such as trajectory deviation, obstacle position, and vehicle attitude, providing a basis for subsequent processing and decision-making;

[0040] 2. Data preprocessing module: Integrates hardware-level low-pass filtering and hardware-fixed Kalman filtering functions to reduce noise and purify the raw data collected by the sensing module, outputting pure state quantities to ensure data reliability;

[0041] 3. CSA-MPC Algorithm Module: Composed of a chaos initialization unit, a simulated annealing optimization unit, an MPC rolling optimization unit, and a fault-tolerant control logic submodule (where CSA is an abbreviation for Chaos-Simulated Annealing), it is responsible for performing algorithmic calculations on the preprocessed data to achieve control parameter optimization and fault-tolerant decision-making;

[0042] 4. Control Module: Includes an industrial-grade PLC, an ARM Cortex-A9 architecture microprocessor, a storage module, and a communication interface board. As the "central hub" of the system, it receives processed data and algorithm results, outputs control commands, synchronizes hardware status, and stores operating data.

[0043] 5. Drive module: It adopts four symmetrically arranged steering wheel mechanisms in the left front / right front / left rear / right rear configuration. Each group is equipped with a drive motor, steering motor, brake and encoder to receive control commands and realize independent drive and steering of the steering wheel.

[0044] 6. Feedback hardware module: Includes 4 encoders and 1 IMU shared with the sensing module. Its core function is to collect feedback data on the actual steering wheel angle, speed and vehicle attitude, providing a basis for closed-loop control.

[0045] 7. Fault Diagnosis Module: Equipped with 4 current sensors, a voltage monitoring module, a temperature sensor, and an alarm unit, it monitors key states such as motor current, power supply voltage, and equipment temperature in real time, detects faults in a timely manner, and triggers early warning or fault-tolerant logic.

[0046] 8. Auxiliary electronic module: including control box cooling fan and status indicator lights, used to ensure stable heat dissipation of the control box and to intuitively display the equipment operating status, respectively;

[0047] 9. Power supply module: Composed of a 48V / 200Ah lithium battery pack, an energy recovery unit and a PMU (power management module), it provides stable power supply to the entire system and recovers braking energy to extend the driving range.

[0048] Figure 1 The document clearly marks the installation location, core parameters, and connection relationships of each module. The thick black solid line represents the main data flow link, the red dashed line represents the fault tolerance priority link, and the blue dashed line represents the energy recovery link. Each link follows the operation rules of data acquisition-processing-decision-execution-feedback, fault detection-early warning-fault tolerance, and energy recovery-storage-reuse to ensure the coordinated and orderly operation of the entire system process.

[0049] (II) Detailed System Design and Relationships Between Modules

[0050] The hardware system in this embodiment adopts an architecture of distributed sensing + centralized control + distributed execution + hierarchical power supply. Each module realizes data interaction and command flow through standardized interfaces and unified communication protocols. The specific composition, parameters, functions and inter-module relationships are as follows:

[0051] 1. Detailed parameters and functions of the hardware module

[0052] Perception Module

[0053] The core components include one LiDAR, two vision sensors, one Inertial Measurement Unit (IMU), and one data fusion board. Key specifications are as follows: LiDAR detection range ≥ 5m, accuracy ≤ 0.01m, detection angle 120°, sampling frequency 100Hz; vision sensor frame rate ≥ 30fps, resolution 1920×1080, minimum obstacle recognition size 0.1m×0.1m; IMU sampling frequency ≥ 100Hz, attitude measurement accuracy ≤ 0.05°, acceleration measurement range ±5g; fusion board supports CAN / Ethernet interface, data processing latency ≤ 0.002s. The core function is to collect trajectory deviation, obstacle position, and vehicle attitude data, and output clean state data after hardware-level filtering and fusion. Installation locations are as follows: the LiDAR is installed at the center of the front of the vehicle roof; the vision sensors are arranged on the left and right sides of the front of the vehicle; the IMU is installed at the geometric center (center of mass) of the vehicle body; and the fusion board is located inside the control box (adjacent to the PLC).

[0054] Control module

[0055] The core components consist of one industrial-grade PLC, one high-performance microprocessor, one storage module, and one communication interface board. Key specifications are as follows: PLC main frequency ≥ 1GHz, memory ≥ 2GB, control cycle ≤ 0.01s, supports CANopen protocol; microprocessor adopts ARM Cortex-A9 architecture, quad-core design, floating-point operation capability ≥ 1GFLOPS, cache ≥ 512KB; storage module capacity ≥ 16GB, read / write speed ≥ 100MB / s; communication interface board supports CAN, Ethernet, and serial port (RS485), adds a multi-AGV collaborative interface, supports IEEE 802.11n wireless communication protocol (transmission rate ≥ 300Mbps, latency ≤ 0.005s), and adds an ETH / Profinet collaborative port. The core functions are to receive sensing data, run the CSA-MPC algorithm and output control commands, synchronize hardware status, store algorithm parameters and operation logs, and realize position synchronization, load distribution and path coordination of multiple AGVs through a multi-AGV collaborative interface, supporting master-slave AGV role switching (the master AGV is responsible for path planning, and the slave AGV receives synchronization commands). The installation location is inside the control box in the middle of the vehicle body (using a dustproof and shockproof installation method).

[0056] driver module

[0057] The core components consist of four symmetrically arranged steering wheel mechanisms (front left / front right / rear left / rear right). Each steering wheel mechanism includes a brushless DC drive motor, a steering motor, an electromagnetic brake, and an incremental encoder. Key specifications are as follows: drive motor rated power ≥ 5kW, rated speed ≥ 3000rpm, torque ≥ 15N·m, speed range 0-1.5m / s; steering motor rated power ≥ 1kW, steering angle range ±45°, steering angle accuracy ≤ 0.1°, response time ≤ 0.02s; brake response time ≤ 0.2s, maximum braking force ≥ 500N; encoder resolution ≥ 1024. The core function is to receive control commands, achieve independent drive and steering of the steering wheels, and provide feedback on actual steering angle and speed data. Installation positions are at the four corners of the vehicle body (front and rear wheel distance L = 2.5m, left and right wheel distance D = 1.8m).

[0058] Power supply module

[0059] The core components include a high-capacity lithium battery pack, an energy recovery unit, a power management module (PMU), and redundant power supply lines. The frame is made of Q235 steel plate. Key specifications are as follows: lithium battery pack voltage 48V, capacity ≥200Ah, cycle life ≥1000 cycles, driving range ≥8h; energy recovery unit recovery efficiency ≥30%, operating voltage range 40-55V; PMU output voltage accuracy ±0.5V, overload protection threshold 1.2 times rated current. Its core functions include powering the entire system, recovering and storing energy during braking, and stabilizing voltage output to ensure high-load operation and execution. It integrates a battery thermal management unit (TMS), containing a temperature sensor (measurement accuracy ±0.3℃), a 50W cooling fan (start threshold 45℃), and an 80W heating element (start threshold 0℃) located inside the battery pack. It monitors battery temperature in real time via a CAN bus, activating the cooling fan when the temperature exceeds 45℃ and the heating element when it falls below 0℃, ensuring the battery operating temperature remains between 10-40℃. This improves battery stability during continuous high-load operation (≥4h) and prevents overheating-induced performance degradation. The installation location is at the bottom of the vehicle body (lower center of gravity for easier load balance).

[0060] Fault diagnosis module

[0061] The core components include four current sensors, a voltage monitoring module, a temperature sensor, and a fault alarm unit. Key specifications are as follows: current sensor measurement range 0-50A, accuracy ≤±1%; temperature sensor measurement range -20~85℃, accuracy ±0.5℃; alarm unit is an audible and visual alarm mode with an alarm volume ≥80dB. The core functions include real-time monitoring of motor current, power supply voltage, and equipment temperature; detection of steering wheel failure / overload faults; and triggering algorithmic fault-tolerant logic. A new remote monitoring interface (supporting 4G / 5G module expansion, data upload frequency 1Hz) allows for remote transmission of equipment status (battery level, steering wheel health, trajectory deviation) and fault information (fault type, occurrence time, location coordinates). The fault alarm unit employs a tiered early warning mechanism: Level 1 warning (current 8-10A / angle deviation 0.3-0.5°) provides only remote notification; Level 2 warning (current >10A / angle deviation >0.5°) activates audible and visual alarms and remote push notifications; and Level 3 warning (steering wheel failure / voltage anomaly) triggers fault-tolerant mode and emergency shutdown, adapting to unattended operation in confined spaces. Installation locations are as follows: the current sensor is connected in series on the steering wheel motor power supply line, with other components located in the control box and key parts of the chassis.

[0062] 2. Inter-module connections and data flow

[0063] In this embodiment, the hardware system adopts a distributed sensing + centralized control + distributed execution + hierarchical power supply architecture. Each module achieves efficient collaboration through physical line connections and standardized communication protocols. The connection relationships and data flow are as follows: Figure 1 The link rules are shown in the annotations (thick black solid lines represent main data flow links, red dashed lines represent fault-tolerant priority links, and blue dashed lines represent energy recovery links), as detailed below:

[0064] (1) Connection between the sensing module and the control module (corresponding) Figure 1 (Main data flow link)

[0065] Physical connection: The LiDAR, vision sensor, and IMU establish a connection with the data fusion board through the CAN bus. The fusion board then connects directly to the microprocessor of the control module through Ethernet, forming a one-way data transmission channel between the sensor, the fusion board, and the microprocessor.

[0066] Communication protocols: The CAN bus uses the CANopen protocol (500kbps baud rate) for transmitting raw sensor data to the fusion board; the Ethernet uses the TCP / IP protocol (100Mbps transmission rate) for transmitting the fused data to the microprocessor.

[0067] Data flow: by Figure 1The main data stream is executed by the perception module, which first collects raw data such as trajectory deviation, obstacle position, and vehicle posture. After hardware-level noise reduction and multi-source data fusion are completed by the fusion board, the pure state quantity is output to the microprocessor. The entire transmission process has a delay of ≤0.002s, ensuring data real-time performance.

[0068] (2) Internal connection of the control module (PLC and microprocessor, corresponding to) Figure 1 (Control center coordination link)

[0069] Physical connection: A dual-link redundancy design of serial port (RS485) + shared memory is adopted to ensure the reliability of data transmission and meet the requirements of real-time interaction.

[0070] Communication protocol: The serial port adopts the Modbus-RTU protocol (baud rate 115200bps) for non-real-time data interaction; the shared memory is used for real-time data transmission, with an access latency of ≤0.001s, which meets the low latency requirements for data interaction within the control module.

[0071] Division of labor and collaboration: The microprocessor, as the core of algorithm operation, runs the CSA-MPC algorithm and outputs control increments; the PLC is responsible for converting control increments into PWM instructions and receiving data transmitted from the feedback hardware. The two work synchronously according to the control cycle (≤0.01s) without timing misalignment, ensuring accurate and timely output of control instructions.

[0072] (3) Connection between the control module and the drive module (corresponding) Figure 1 Execution Link

[0073] Physical connection: The PLC is connected to the drive motor and steering motor controller of the four steering wheels one by one through the PWM output interface. At the same time, the encoder equipped on the steering wheel establishes a feedback connection with the PLC through the CAN bus, forming a closed loop link of control command output and execution status feedback.

[0074] Signal type and parameters: The control signal is a PWM signal (duty cycle 0~100%), used to precisely adjust the motor speed and direction angle; the feedback signal is the encoder A / B / Z phase signal;

[0075] Drive coordination logic: The PLC distributes control commands in the order of "left front → right front → left rear → right rear". The steering motor and drive motor track the target value through closed-loop control to ensure that the angle deviation is ≤0.1° and the speed deviation is ≤0.01m / s, and to avoid the steering wheel sliding friction.

[0076] (4) Connection between the power supply module and other modules (corresponding) Figure 1 (Power supply link and energy recovery link)

[0077] Physical connection: The PMU provides graded power supply (12V / 24V / 48V) to each module through redundant power supply lines. The lines are equipped with fuses, shielding layers and protective interfaces to ensure power supply safety. The energy recovery unit is connected to the lithium battery pack through the blue dashed line link to realize unidirectional recovery of braking energy.

[0078] Power supply distribution and characteristics: The power supply line is strictly matched to the power requirements of each module. Key components such as the control module and drive module adopt a dual-line redundancy design to avoid power outages caused by single-point failures. The energy recovery unit recovers electrical energy during braking and stores it in the lithium battery pack to prevent reverse flow of electrical energy from affecting system stability.

[0079] (5) Connection between the fault diagnosis module and other modules (corresponding) Figure 1 (Fault-tolerant link)

[0080] Physical connection: The current sensor is connected in series on the power supply line of the steering wheel motor, the voltage monitoring module is connected to the power supply bus, and the temperature sensor is arranged in the control box and key parts of the frame. Each sensor monitors the operating status of the corresponding hardware and establishes a data transmission channel with the PLC through the CAN bus.

[0081] Data flow and fault response: By Figure 1 The red dashed line indicates the fault-tolerant link execution. Sensors collect data such as motor current, supply voltage, and equipment temperature in real time and transmit them to the PLC. When the detected data exceeds the fault threshold, the hardware immediately triggers a power outage and audible and visual alarm. Simultaneously, the fault-tolerant control mode of the CSA-MPC algorithm is activated at the algorithm level, reallocating control commands to healthy components to ensure stable system operation. For example, when a steering wheel failure is diagnosed, the fault-tolerant control logic re-solves the inverse kinematics based on the geometric position of the remaining healthy steering wheels, appropriately relaxing the speed and angle constraints on the failed steering wheel, and recalculating the control commands using the CSA-MPC algorithm.

[0082] (III) Construction and Working Principle of Dynamic Model

[0083] 1. Basis for Model Establishment

[0084] To accurately describe the motion of high-load AGVs, this invention constructs the following collaborative sub-models: Based on nonholonomic constraint theory and Ackermann steering principle, this embodiment virtualizes the four-steering wheel model into a two-steering wheel model. Through longitudinal and lateral dynamic decoupling, it describes the load resistance and attitude change characteristics respectively, accurately adapting to the motion law of 5-ton high-load scenarios. The model parameters are fully matched with the hardware system.

[0085] 2. Core Model Formula and Explanation

[0086] (1) Lateral two-degree-of-freedom dynamic model (describing attitude change)

[0087]

[0088] Where: m is the total mass of the vehicle body (including a 5-ton load, m = 8000 kg); u is the longitudinal speed; It is the centroid sideslip angle; This refers to the yaw rate; , These are the lateral forces of the front and rear wheels, respectively. The moment of inertia about the Z-axis (which can be calculated / experimentally measured based on the geometric distribution of the vehicle body and load; in this embodiment, it is taken as...) =2000 kg·m 2 ); a is the distance from the front wheel to the center of gravity (a=1.2m); b is the distance from the rear wheel to the center of gravity (b=1.3m).

[0089] Working principle: It is assumed that the tire lateral force is linearly related to the lateral angle. , For lateral stiffness, =50000N / rad), accurately calculate the attitude change caused by the center of gravity shift under high load, and provide a basis for error compensation for the control algorithm.

[0090] (2) Longitudinal dynamic model (describing load resistance)

[0091]

[0092] in: This represents the total longitudinal resistance. ( (Longitudinal acceleration) and (Acceleration drag) ( The air drag coefficient, ; For windward area, m 2 ; air density, kg / m 3 () represents air resistance; ( (Road slope) represents the slope resistance; ( The rolling resistance coefficient, =0.02) is the rolling resistance.

[0093] Working principle: The longitudinal resistance under different working conditions is calculated in real time to provide constraints for the speed optimization of the MPC algorithm and ensure the power output matching requirements under high load.

[0094] (3) Calculation model of four steering wheel speed and deflection angle (based on Ackermann principle)

[0095]

[0096] in: The speed of the four steering wheels; The deflection angle of the four steering wheels; It is the turning angular velocity; The turning radius of the four steering wheels; For instantaneous velocity; The center point of the steering wheel.

[0097] Working principle: The instantaneous center of velocity is calculated by a virtual dual steering wheel model. Combined with the Ackermann steering principle, the motion of the four steering wheels is consistent, avoiding sliding friction and improving motion stability and steering wheel life.

[0098] (4) Vibration compensation sub-model (adapted to uneven road surface interference)

[0099]

[0100] in: The vibration interference force caused by uneven road surface; The virtual suspension equivalent stiffness (consistent with the virtual suspension parameters of the mechanical structure) =50N / mm)); For virtual suspension equivalent damping ( z is the vertical displacement (obtained by integrating the vertical acceleration collected by the IMU); This represents the vertical velocity.

[0101] Working principle: The IMU collects vertical acceleration and vibration frequency in real time, and calculates vibration disturbance force by combining virtual suspension parameters. This force is used as a compensation term in the longitudinal / lateral dynamic model to correct the steering wheel speed and angle commands, and to counteract the trajectory deviation caused by uneven road surfaces.

[0102] (iv) CSA Optimized MPC Control Algorithm Design and Working Principle

[0103] 1. Overall Algorithm Architecture

[0104] The algorithm consists of three layers: the bottom layer is the dynamic model layer (providing the basis for state prediction); the middle layer is the MPC core layer (rolling optimization of control increments); and the top layer is the CSA optimization layer (optimizing MPC parameters and avoiding local optima). The three layers work together to achieve high-precision and high-real-time control.

[0105] 2. Core Algorithm Steps and Working Principle

[0106] (1) MPC core layer design

[0107] Prediction Time Domain and Control Time Domain: Dynamic Configuration When turning in a narrow space During linear acceleration To balance prediction accuracy and computational efficiency.

[0108] Objective function:

[0109]

[0110] in: This is the system state vector (including position, velocity, and attitude). The reference state vector; To control the increment vector; Q and R are weight matrices (Q=diag [10,10,5], R=diag [2,2]); For relaxation factor ( ); As a fault-tolerance compensation term, and as a preferred option, it can be understood as a vector of relaxation variables, the dimensions of which correspond to the state constraints. When the fault diagnosis module detects that the performance of an actuator is limited, the corresponding constraint... Elements can be increased to soften constraints and ensure that the optimization problem has a solution.

[0111] Constraints: Steering wheel deflection angle range [-45°, 45°], speed range [0, 1.5 m / s], acceleration range [-0.5 m / s²]. 2 0.5m / s 2 ].

[0112] Working principle: Through rolling time-domain optimization, the objective function is solved once every 0.01s, and the optimal control increment ΔU* is output to correct the steering wheel control command in real time to deal with dynamic disturbances.

[0113] (2) CSA optimization layer design (optimizing the Q and R matrices of MPC and , parameter)

[0114] Step 1: Chaotic Initialization (Logistic Mapping)

[0115] , ,

[0116] Generate 50 chaotic sequences and map them to feasible solutions for the MPC parameters (Q and R matrix elements range [1, 20]). , This increases the diversity of initial solutions and avoids getting trapped in local optima.

[0117] Step 2: Simulated Annealing Optimization

[0118] Initial temperature: ,in , The objective function extrema of the initial solution. =0.9 is the initial acceptance probability.

[0119] Metropolis Criterion: Probability of Acceptance of New Solution , The objective function difference between the new solution and the current solution, T, decreases with the number of iterations. , ).

[0120] Iteration termination condition: number of iterations Rate of change of the objective function .

[0121] Working principle: By leveraging the ergodicity of chaotic sequences and the global search capability of simulated annealing, MPC parameters are adaptively optimized, enabling MPC to achieve optimal control performance under different operating conditions (acceleration, steering, obstacle avoidance, complex path tracking), thus solving the problems of traditional MPC parameters relying on experience and lacking robustness.

[0122] (3) Algorithm execution flow

[0123] Reference Figure 1 The main data flow and feedback loop are shown below:

[0124] 1. The perception module collects trajectory deviation, obstacle position, and vehicle attitude data, and transmits them to the control module;

[0125] 2. The dynamic model updates the predicted values ​​of longitudinal drag and lateral attitude in real time based on the collected data;

[0126] 3. The CSA optimization layer optimizes the Q and R matrices of MPC based on the current operating conditions. , parameter;

[0127] 4. The MPC core layer uses the optimized parameters to solve the rolling time-domain optimization problem and outputs the optimal control increment. ;

[0128] 5. Control increments are converted into the deflection angles of the four steering wheels. With speed The command drives the steering wheel to execute the command;

[0129] 6. The hardware collects the actual execution results and transmits them to the control module, which drives the algorithm to correct parameters and form a closed-loop control.

[0130] (4) Adaptive compensation steps for steering wheel wear

[0131] 1. Cumulative value of angular deviation based on encoder feedback Establish a wear assessment model based on the running time t: The wear coefficient is... W is the time decay coefficient, and W is the wear level, ranging from [0,1].

[0132] 2. When At that time, the compensation logic is activated: the weights of the Q matrix of MPC are adjusted through the CSA algorithm (the lateral deviation weight is increased from 10 to 15), and the steering wheel angle command is corrected. , (for compensation coefficients)

[0133] 3. The hardware provides real-time monitoring of the compensation effect. If the angle deviation is still >0.1°, a mechanical adjustment prompt is triggered (status indicator light flashes). The mechanical clearance is then finely adjusted through the adjustment hole of the steering wheel mounting base to ensure the angle accuracy after long-term operation under high load (≥500h).

[0134] (v) Simulation verification and engineering adaptation

[0135] This embodiment demonstrates a simulation test of the systematic solution designed above:

[0136] 1. Simulation Platform and Scene Design

[0137] Platform: MATLAB / Simulink, building a hardware-in-the-loop simulation model, integrating four-steering wheel dynamics, sensor noise model, and environmental interference model (road slope ±5°, sudden obstacles).

[0138] Scenario: Straight-line acceleration (0→1.5m / s), turning (turning radius 2.5m), obstacle avoidance (obstacles appear suddenly at a distance of 0.5m), complex path tracking (sine wave trajectory, amplitude 0.5m, period 5m).

[0139] 2. Simulation Results (Hardware-Algorithm Fusion Effect)

[0140]

[0141] 3. Engineering Adaptation Measures

[0142] Lightweight algorithm: The CSA-MPC algorithm is optimized by replacing floating-point operations with fixed-point operations, and the single operation time is ≤0.005s, which is adapted to the computing power of microprocessors;

[0143] Hardware redundancy design: The steering wheel mechanism adopts dual motor backup, the sensing module adopts dual redundancy of LiDAR and vision sensor, and the power supply module adopts dual line redundancy to improve system reliability;

[0144] Ease of installation and commissioning: The control box is designed with a modular structure, and each hardware module can be installed independently and is easy to disassemble; the steering wheel mounting base is reserved with adjustment holes, which can finely adjust the wheel track and wheelbase to ensure consistency with the dynamic model parameters.

[0145] The solution provided in this embodiment addresses the stringent requirements of high-load and confined space handling for high-voltage cables. It constructs a comprehensive collaborative system encompassing hardware support, algorithmic decision-making, data closed-loop processing, and energy assurance. Through deep integration of hardware and software and core technological innovation, it achieves a comprehensive breakthrough in handling performance and practical value. The solution innovatively integrates the Chaos-Simulated Annealing (CSA) algorithm with Model Predictive Control (MPC). It generates diverse initial solutions through Logistic chaotic mapping and optimizes the weight matrix and prediction / control time-domain parameters of MPC using the global search capability of the simulated annealing algorithm. This dynamically adapts to different operating conditions such as linear acceleration and turning in confined spaces, effectively avoiding the pitfalls of traditional control algorithms that easily get trapped in local optima. Combined with the dual-redundancy configuration of the LiDAR and visual sensors in the perception module, and hardware-level and software-level dual filtering processing, the average trajectory deviation for complex path tracking is reduced to 0.03m, accurately adapting to the handling requirements of complex trajectories such as sine waves.

[0146] For high-load scenarios involving 5 tons, the design incorporates a dynamic model that combines longitudinal drag compensation with lateral two-degree-of-freedom decoupling. This model accurately characterizes the impact of center-of-gravity shift and load drag on the motion state. Combined with a symmetrical four-steering-wheel layout (2.5m front-to-rear wheel spacing and 1.8m left-to-right wheel spacing) and a virtual suspension structure, closed-loop control achieves precise coordination between steering wheel speed and yaw angle. This avoids sliding friction while controlling the maximum lateral acceleration during steering to 1.2m / s². 2 Within this range, the vibration amplitude during dynamic motion is reduced by 30%, effectively suppressing the risk of rollover and trajectory deviation under high loads. Meanwhile, after lightweight optimization, the algorithm's single computation time is ≤0.005s, and the control cycle is shortened to 0.01s. Combined with the high-speed TCP / IP communication protocol (100Mbps), the obstacle avoidance reaction time is reduced to 0.8s, the path deviation distance is only 0.5m, and the recovery time is 1.0s, perfectly meeting the need for rapid avoidance of sudden obstacles in confined spaces.

[0147] The solution's end-to-end fault tolerance mechanism and dynamic adaptive capabilities further enhance operational reliability in complex environments. A fault diagnosis system comprised of current sensors and voltage monitoring modules establishes a tiered early warning mechanism, enabling rapid response to faults such as steering wheel failure and voltage anomalies. The CSA-MPC algorithm reallocates control commands to healthy steering wheels, while simultaneously addressing dynamic disturbances such as road slope variations of ±5° and changes in air resistance, ensuring an average trajectory deviation fluctuation of ≤0.01m, significantly outperforming traditional solutions. Furthermore, the independent design of the tiered power supply architecture (12V / 24V / 48V) and the regenerative braking link achieves an energy recovery efficiency of ≥30%, increasing the driving range by over 25% compared to traditional solutions and meeting the requirements for long-distance transport of high-voltage cables. Modular control box design, pre-drilled adjustment holes in the steering wheel mounting base, and optimized algorithm fixed-point computation, among other engineering adaptation measures, facilitate hardware procurement and maintenance, and allow the algorithm to run on conventional industrial controllers, significantly reducing implementation difficulty.

[0148] The overall solution of this invention achieves a leapfrog improvement in core performance such as trajectory accuracy, response speed, and stability through deep collaboration of hardware layout parameters (such as steering wheel spacing and sensor installation position), core model parameters (such as mass and moment of inertia of the dynamic model), key algorithm formulas (such as chaotic sequence generation and simulated annealing temperature calculation), and control architecture (such as data transmission protocol and control cycle). Furthermore, it ensures long-term operational reliability and engineering practicality through standardized and redundant design, providing an efficient and safe solution for the high-load handling of high-voltage cables in narrow and confined spaces such as substations and underground utility tunnels.

[0149] Based on the same inventive concept, this invention also provides a computer device, comprising: one or more processors, and a memory for storing one or more computer programs; the programs include program instructions, and the processor executes the program instructions stored in the memory. The processor may be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. It is the computing and control core of the terminal, used to implement one or more instructions, specifically for loading and executing one or more instructions stored in a computer storage medium to implement the above-described method.

[0150] It should be further explained that, based on the same inventive concept, the present invention also provides a computer storage medium storing a computer program, which, when executed by a processor, performs the above-described method. This storage medium can be any combination of one or more computer-readable media. A computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. A computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of computer-readable storage media (a non-exhaustive list) include: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In the present invention, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0151] In the description of this specification, references to terms such as "an embodiment," "example," "specific example," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this disclosure. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0152] The foregoing has shown and described the basic principles, main features, and advantages of this disclosure. Those skilled in the art should understand that this disclosure is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of this disclosure. Various changes and modifications can be made to this disclosure without departing from its spirit and scope, and all such changes and modifications fall within the scope of this disclosure as claimed.

[0153] This invention is not limited to the preferred embodiment described above. Anyone inspired by this invention can derive various other methods and systems for trajectory tracking control of high-load omnidirectional AGVs. All equivalent variations and modifications made within the scope of the claims of this invention shall fall within the scope of this invention.

Claims

1. A trajectory tracking control method for a high-load omnidirectional AGV, wherein the omnidirectional AGV is driven by multiple independently steering and driving steering wheel mechanisms, characterized in that, include: Real-time acquisition of vehicle status information and desired trajectory information of the omnidirectional AGV; Based on the vehicle state information and the desired trajectory information, the core parameters of the model predictive controller are optimized online using the chaotic simulated annealing algorithm. The core parameters include the state weight matrix, the control increment weight matrix, the prediction time domain, and the control time domain. The model predictive controller performs state prediction based on a longitudinal-lateral decoupled dynamic model adapted to high-load conditions. Using the optimized core parameters, the model predicts the controller to perform rolling time-domain optimization, and the optimal control increment is obtained by solving the problem. The optimal control increment is converted into a drive command for the multi-steering wheel mechanism, controlling the omnidirectional AGV to track the desired trajectory.

2. The trajectory tracking control method for a high-load omnidirectional AGV according to claim 1, characterized in that: The longitudinal-lateral decoupled dynamic model includes a lateral two-degree-of-freedom model for describing changes in vehicle attitude, a longitudinal resistance model for calculating load resistance under high load conditions, and a vibration disturbance compensation sub-model. The longitudinal resistance model covers acceleration resistance, air resistance, slope resistance, and rolling resistance. The vibration disturbance compensation sub-model calculates road disturbance force based on the equivalent stiffness and equivalent damping of the virtual suspension, combined with real-time acquired vertical displacement and vertical velocity, and outputs it as a compensation term to correct the model.

3. The trajectory tracking control method for a high-load omnidirectional AGV according to claim 1, characterized in that: The optimization process of the chaotic simulated annealing algorithm includes: generating a chaotic sequence using a Logistic mapping and mapping it to a feasible solution for the core parameters of the model predictive controller; calculating the initial temperature based on the extremum of the objective function of the initial solution and the preset initial acceptance probability; determining whether a new solution is acceptable using the Metropolis criterion; decreasing the temperature with the number of iterations according to a preset decay coefficient; and outputting the optimal core parameters when the number of iterations reaches a preset iteration threshold or the rate of change of the objective function meets the convergence condition. The prediction time domain and control time domain of the model predictive controller are dynamically configured according to the operating conditions, and its objective function includes a state tracking error term, a control increment constraint term, and a fault-tolerant relaxation term.

4. The trajectory tracking control method for a high-load omnidirectional AGV according to claim 1, characterized in that: The process of converting the optimal control increment into drive commands includes: based on the Ackermann steering principle, calculating the instantaneous center of velocity through virtual dual steering wheel modeling, deriving the target deflection angle and drive speed of each steering wheel, ensuring the motion consistency of multiple sets of symmetrically arranged steering wheels, and avoiding sliding friction.

5. The trajectory tracking control method for a high-load omnidirectional AGV according to claim 1, characterized in that: It also includes adaptive compensation and fault tolerance steps for steering wheel wear: a wear assessment model is established based on the cumulative value of the angle deviation fed back by the encoder and the running time. When the wear level exceeds or reaches the preset wear level threshold, the state weight matrix of the model prediction controller is adjusted and the steering wheel deflection angle command is corrected. The system monitors the steering wheel motor current, power supply voltage, and equipment temperature in real time. When the monitored data exceeds or reaches the preset monitoring threshold, a graded early warning mechanism is triggered. When steering wheel failure or voltage abnormality is detected, the fault-tolerant mode is activated, and control commands for the healthy steering wheel are reallocated.

6. The trajectory tracking control method for a high-load omnidirectional AGV according to claim 1, characterized in that: The control module of the omnidirectional AGV supports a multi-AGV collaborative interface, and realizes the position synchronization, load distribution and path coordination of multiple AGVs through a wireless communication protocol. It supports master-slave AGV role switching, with the master AGV responsible for path planning and the slave AGV receiving synchronization control commands.

7. A trajectory tracking control system for a high-load omnidirectional AGV, wherein the omnidirectional AGV is driven by multiple independently steering and driving steering wheel mechanisms, characterized in that, include: The perception module is used to collect the vehicle status information and desired trajectory information of the omnidirectional AGV in real time. The control module is configured to optimize the core parameters of the model predictive controller online using a chaotic simulated annealing algorithm based on the vehicle state information and the desired trajectory information. The core parameters include a state weight matrix, a control increment weight matrix, a prediction time domain, and a control time domain. The model predictive controller performs state prediction based on a longitudinal-lateral decoupled dynamic model adapted to high-load conditions, and then performs rolling time domain optimization using the optimized core parameters to solve for the optimal control increment. The drive module is used to receive the optimal control increment output by the control module, convert it into drive instructions and execute them to control the omnidirectional AGV to track the desired trajectory. The power supply module is used to provide stable power to the sensing module, control module and drive module; The fault diagnosis module is used to monitor the operating status of each module in real time.

8. The trajectory tracking control system for a high-load omnidirectional AGV according to claim 7, characterized in that: The control module executes the trajectory tracking control method as described in claim 1 to achieve a closed-loop control process encompassing vehicle state information acquisition, model prediction controller parameter optimization, optimal control increment solution, and drive command generation.

9. The trajectory tracking control system for a high-load omnidirectional AGV according to claim 7, characterized in that: The perception module collects vehicle status information through lidar, vision sensors, and inertial measurement units. The collected data is preprocessed by hardware-level low-pass filtering and Kalman filtering before being output. The perception module and control module adopt a dual redundancy design.

10. The trajectory tracking control system for a high-load omnidirectional AGV according to claim 7, characterized in that: The power supply module adopts a multi-voltage level hierarchical power supply architecture, with dual-line redundancy configured for key modules. During braking, electrical energy is recovered through an energy recovery unit. The power supply module also integrates a thermal management unit to maintain stable operating temperature.

Images